10 ways practice administrators can put AI to work

For many medical practices, artificial intelligence (AI) is no longer a hypothetical. It is sitting inside the electronic health record (EHR), powering ambient documentation tools and increasingly surfacing in revenue cycle platforms and contact centers.

A January 2025 Medical Group Management Association (MGMA) Stat poll found that AI tools were the top technology focus for health care organizations, with 32% of respondents ranking them their leading priority — ahead of electronic health record usability (30%) and revenue cycle management systems (17%).

The same feature noted that approximately one-third of U.S. health care organizations had integrated AI into their workflows and that more than 70% of those using AI said they were seeing at least some success.

That interest is now showing up in practice-level surveys. A company survey of 887 health care professionals conducted by eClinicalWorks and reported by Medical Economics found that half of practices are already using at least one AI tool. Among AI adopters in that sample, 72% relied on Sunoh.ai for clinical documentation support, and about two-thirds of those users said it saved them one to four hours or more per day.

Fax workflows, another longstanding pain point, were also being reshaped: about 64% of Image AI users said automated fax sorting and processing saved them at least an hour a day.

AI is also moving deeper into clinical support. A 2025 athenahealth survey polled 501 physicians and practice administrators. Among respondents who used AI, 60% said it helped them look up clinical information, 55% said it brought lab and imaging results into a single view, and 56% said it surfaced recent clinical evidence during patient visits.

Sixty percent preferred AI as an assistant to review charts and catch small details, while 26% said they were comfortable fully delegating that specific chart-review task to AI; in total, 86% were comfortable either fully delegating or having AI assist with catching details across patient records.

Still, adoption is uneven. A 2024 American Medical Association (AMA) study cited in a Physicians Practice article by Gary Wietecha, M.D., found that 38% of doctors were using AI features, mostly for documentation and care plans, and 43% of those users found the tools beneficial. But a separate HIMSS study he highlighted reported that only 27% of staff members viewed AI positively.

That gap helps explain why practices can buy AI-enabled tools and still see lukewarm results.

Financial pressure is another driver. The AMA has warned that decreasing payment, high-cost resources and administrative burdens are destabilizing independent private practice and pushing physicians toward hospital and private equity ownership.

Julie Lambert, now senior director of health industry group at Alvarez & Marsal, wrote in Medical Economics that the AMA “cites challenges in managing payment as a top reason that practices are forced to sell or go out of business,” alongside staff shortages and burnout.

Her piece drew on an Inovalon survey showing that eligibility, prior authorization, denials and patient payments are among the most significant revenue cycle pain points where AI could help.

Denials data tell a similar story. An article by Neal K. Shah in Medical Economics noted that out of roughly $3 trillion in claims submitted annually, about $262 billion are initially denied, citing external claims data.

That same article pointed to research indicating that physicians and staff can spend 10 or more hours a week working denials, and that only about 1% of patients appeal them — even though more than half of those appeals succeed.

Clarissa Riggins of Experian Health told Medical Economics that denial rates have climbed in recent years, and that AI-driven platforms can review documentation in real time, suggest correct codes and flag likely denials before claims go out the door.

Against that backdrop, practice administrators are being asked to turn AI from a buzzword into a practical management tool.